Not looking at your design , I can see your slew rate limiting is unequal.

Each polarity of push-pull driver may have a different impedance and thus the asymmetry causes even harmonic distortion to you added to anticipated odd harmonics.

You should realize the GBW requirements increase dramatically when you include specs for THD at a signal in the same decade as your BW, since you have very little excess gain to feedback for linear error correction.

Also it is important to remember that when designing sharp or high Q RC active filters this also raises the GBW requirements significantly, much higher than you may expect due to phase shifting 1 decade below breakpoint becomes more critical with higher Q filters like Chebychev vs Bessel. TI's Active Filter designer tells GBW requirements for each stage.

Not looking at your design , I can see your 1) slew rate limiting AND/OR 2) hFE is unequal.

1. Each polarity of push-pull driver may have a different impedance and thus the asymmetry causes even harmonic distortion to you added to anticipated odd harmonics.

-You should realize the GBW requirements increase dramatically when you include specs for THD at a signal in the same decade as your BW, since you have very little excess gain to feedback for linear error correction.

Also it is important to remember that when designing sharp or high Q RC active filters this also raises the GBW requirements significantly, much higher than you may expect due to phase shifting 1 decade below breakpoint becomes more critical with higher Q filters like Chebychev vs Bessel. TI's Active Filter designer tells GBW requirements for each stage.

. 2. Unequal current gain is another reality near current limits for high voltage outputs. This affects the THD by adding even & odd harmonics.

e.g.

(Disregard the frequency, I adjust the harmonic levels to resemble yours)